
In the residual analyses, we can examine our models in terms of the behaviour of its
residuals. We see that the residuals have a constant mean and variance, which is highly
required in a good time series model. In the second chart, residuals clearly show a
normal distribution pattern with no visible skew or autocorrelation. These findings
strengthen our belief in the accuracy of the final model, and show us that there is little to
no remaining unaccounted for information that is required for the model.
Conclusions:
In this study, we addressed the challenge of forecasting the hourly solar power
production at the Edikli Solar Power Plant in Niğde, Turkey. Given that solar energy
production is inherently dependent on meteorological factors, we aimed to develop a
model that leverages these variables to predict solar power generation accurately.Our
approach involved several stages. First, we organized our data in Python and applied a
general regression analysis. After this initial step, we decided to continue our work in R.
We then applied Lasso regression to select the most significant variables, refining our
feature set. We evaluated models with different sets of lagged values (Lag 10, Lag 24,
and Lag 36) and lambda values. Next, we developed ARIMA models for each hour and
an overall ARIMA model, accounting for temporal dependencies and seasonality. The
best-performing approach was a linear regression model with Lasso-selected variables,
a transformed hour, and lagged production values (Lag 24 and Lag 36). The resulting
model had the best performance metrics and showed the highest forecasting accuracy.
and thorough residual analysis showing no significant autocorrelation or non-normality.
In conclusion, our study highlights the importance of incorporating meteorological
variables, appropriate feature transformations, and lagged production values in
developing effective forecasting models for solar power production. The performance of
the final model underscores its potential for real-world applications in electricity pricing
and capacity planning, providing a valuable tool for optimizing operations at the Edikli
Solar Power Plant.